For many years the occurrence of disease-specific volatiles have been used as a supporting factor in the clinical diagnoses of various disorders, e.g., the sulfide emission in the breath of Helicobacter pylori-positive patients with gastrointestinal issues (Hoshi et al., 2002); and the sweaty feet odor of patients with isovaleric acidemia (Budd et al., 1967). As a result, the study of volatile organic compounds (VOCs) produced by commensal and pathogenic microorganisms has emerged as a path to characterizing these disease-specific volatiles. In particular, the last 15 years has seen a significant rise in the study of microbial VOCs due to the universal implementation of improved analytical methodology and data analysis techniques. Comprehensive sampling and analytical methods have broadened the spectrum of compounds that can be investigated while the incorporation of dimension reduction and clustering methods has enabled the identification of discriminatory trends across the microbial VOC data. Most studies have been primarily focused on the investigation of in vitro microbial cultures and have been critical in identifying metabolic and cellular pathways of particular compounds. These studies have demonstrated that species-level differences in VOC production do exist between pathogenic and commensal microbial species and highlight the need for further study.

The diversity and mechanisms behind microbial volatilomes have been recently illustrated in several comprehensive review papers (Schulz and Dickschat, 2007; Elmassry and Piechulla, 2020; Kai, 2020; Weisskopf et al., 2021) and books (Ryu et al., 2020). Growth parameters such as growth media (Jenkins and Bean, 2019, 2020), growth phase (Filipiak et al., 2012b; Chen et al., 2016), oxygen content (Insam and Seewald, 2010), temperature (Misztal et al., 2018) and pH (Schulz-Bohm et al., 2017), all influence microbial volatilomes. Another less studied factor in the overall variation seen across the volatilomes of microbial species is the occurrence of strain-level specificity in volatilomic emission within a given species (Schulz et al., 2020; Weisskopf et al., 2021). In addition to this, the variation in sampling [SPME (Timm et al., 2018; Fitzgerald et al., 2020), thermal desorption tubes (Filipiak et al., 2012a,b), direct syringe (Ashrafi et al., 2018a,b)] and analytical techniques [gas chromatography–mass spectrometry (GC-MS) (Filipiak et al., 2012b; Timm et al., 2018; Fitzgerald et al., 2020), selected ion-flow-tube (SIFT) MS (Shestivska et al., 2015), proton transfer reaction (PTR) MS (Bunge et al., 2008)] employed across the field also have a direct influence on the VOC profiles reported in the literature. Consequently, cross-study validation of reported microbial VOC profiles remains a major challenge in the field. However, the recent establishment of the mVOC 2.0 database (Lemfack et al., 2017) has allowed for some qualitative comparisons of microbial volatilomes and will evolve to be a valuable tool in the field. The database contains thousands of logged compounds along with growth conditions and analytical methods used to acquire the volatilomes of a wide range of microbes. With such a platform available, a community-wide effort is required to build on it and to ensure that the microbial volatilomic profiles available on the database are as comprehensive as possible with respect to the literature. This will ultimately allow the full examination of individual microbial volatilomic profiles relative to all the conditions in which they have been previously examined. Therefore, in order to elucidate the full spectrum of microbial volatilomes, there is a strong need to investigate them under varying conditions in controlled settings.

Staphylococcus aureus, P. aeruginosa, and E. coli are highly prevalent wound pathogens and are responsible for particularly severe infections in diabetic foot ulcers (DFUs) (Gardner et al., 2012; Ndosi et al., 2017). It is estimated that around one in four people with diabetes will develop a diabetic foot ulcer (DFU) in their lifetime (Armstrong et al., 2017). Infections of DFUs highly increase the risk of poor outcomes such as amputation (Abdulrazak et al., 2005). The duration of the DFU is proportional to its severity and is closely associated with species- and strain-level microbial diversity within the infection (Abdulrazak et al., 2005; Kalan et al., 2019). Currently in clinics, time consuming techniques such as blood tests and traditional plate-based techniques are employed to detect potential infections (Lipsky et al., 2019). As early detection is critical in preventing severe infections, rapid non-invasive detection of pathogenic bacterial volatiles in wounds and wound samples could potentially speed up the turnover of clinical information and greatly contribute to the clinical workflow. Our group is currently working on detecting pathogen-specific volatile compounds in DFU swab samples. Volatilomic profiling of pure microbial cultures has played a critical role in our preparation for clinical volatilomic work and also for the interpretation of the data obtained.

The stability of bacterial volatilomes both in different nutritional conditions and strain-to-strain remains relatively understudied. In this study, we examined multiple strains of E. coli, P. aeruginosa, and S. aureus volatilomes across different growth media. By examining microbial volatilomic variability across different strains and different media, the core and accessory volatilomes of these bacteria can be elucidated. These terms were introduced by Bean et al. (2016) but for the context of this study, core compounds refer to compounds that are emitted by both strains across all media; accessory compounds are compounds emitted by at least one strain in at least one medium. Our key objectives of this work were (1) to obtain comprehensive volatilomic data for multiple strains of S. aureus, P. aeruginosa, and E. coli in three different growth media (BHI, LB, TSB); (2) to assess the stability and variation of the observed bacterial volatilomes; and (3) to temporally investigate strain-level specificity within the selected volatilomes by comparing the emission of specific compounds at progressive stages of growth and development of the cells.

Microorganisms produce a diverse range of volatile metabolites that have different physico-chemical properties and biological activities. These volatile metabolites serve an important role in inter-species and inter-kingdom communication and are involved in both beneficial and deleterious interactions between microorganisms (Hoshi et al., 2002). Microbial VOCs have been proposed as potential biomarkers of disease (Elmassry and Piechulla, 2020) that can be non-invasively analyzed to support clinical workflows in the future. Clinical studies involving untargeted breath profiling analyses have demonstrated the potential discriminatory power and diagnostic potential of VOCs for diseases such as pneumonia (Schnabel et al., 2015), cystic fibrosis (Robroeks et al., 2010; Neerincx et al., 2016), tuberculosis (Zetola et al., 2017), and COVID-19 (Ruszkiewicz et al., 2020). It has also been well-established that microbial diversity at the species-level is associated with poor outcomes and longer durations of diabetic ulcer wounds (Abdulrazak et al., 2005). A recent study also reported that diversity at the strain-level of specific pathogenic species is a large contributor to infection severity (Kalan et al., 2019). Consequently, the development of a non-invasive rapid sampling platform potentially capable of discriminating microbes at the species- and strain-level is highly desirable in clinic settings. In the conclusion of our previous study (Fitzgerald et al., 2020), we outlined that our future work will involve the volatilomic analysis of wound samples to identify infection-specific markers. Early unpublished results of this work suggests that volatile markers detected in this study such as 3-methylbutyric acid, acetic acid, 3-methyl-1-butanol, propanoic acid and ethanol do heavily persist in samples obtained from severe wound infections. However, this work is currently ongoing and a higher number of samples are required from a varying spectrum of infected wounds to further support these early findings.

Volatilomic profiling of pure microbial cultures has played a critical role in identifying the cellular origins of metabolites associated with specific species and strains of pathogens. In this study we employed HS-SPME-GC-MS to investigate the stability of the volatilome of multiple strains of prominent wound-associated pathogens across different nutrient-rich growth media. Oxygen availability is expected to have a strong influence on volatile compound formation by bacteria as it dictates whether respiration or fermentation pathways are utilized for metabolism (Weisskopf et al., 2021). Each bacterial sample was made up to a total volume of 5 mL, leaving a 15 mL headspace above the culture. It is predicted that the bacteria utilized the available oxygen in this headspace to carry out aerobic respiration in the early stages of growth. The examined pathogens are capable of anaerobic and/or fermentation under specific conditions, and may be capable of shifting their metabolism from aerobic respiration in oxygen-exhausted conditions (Lemfack et al., 2020). It was expected that this occurred during the 24 h incubation and consequently the volatilomes reported in this study are the result of both aerobic and anaerobic metabolism. Strain-level differences that were observed in the examined bacterial volatilomes were further investigated by analyzing and comparing the volatilome of each strain at progressive phases of cell growth and development.

Dimension reduction methods such as principal component analysis (PCA) (Tait et al., 2013; Chen et al., 2016; Fitzgerald et al., 2020) and clustering methods such as hierarchical clustering (HC) (Bean et al., 2016; Fitzgerald et al., 2020) are frequently employed to identify species-specific trends across the bacterial volatilomic samples. PCA identifies species- and strain-specific volatilomic differences across the bacterial volatilomes and amplifies these differences by constructing new linear variables called principal components (PCs), along which the variation is maximal. The PCs can then be visualized using scores plots. Scores plots show inter-sample distances that clearly illustrate patterns across the data which can be used to identify groups that characterize the overall dataset (Ringnér, 2008). In this study we used these techniques to perform an unsupervised analysis of the volatiomes obtained from the examined strains across each growth media. The PCA scores plots shown in Figure 1 illustrate that across the media the examined bacterial volatilomes demonstrated similar differentiation from each other which allowed clear and consistent species-level discrimination. These scores plots also indicate a certain degree of stability of the whole volatilomes of the examined species across the media. Similar volatilomic stability across different media was previously demonstrated by Dryahina et al. (2016)—in this study, PCA analysis of SIFT-MS volatilomic data showed that S. aureus and P. aeruginosa samples cultured in BHI, NB and Mueller Hinton Broth (MHB) exhibited species-dependent clustering rather than media-dependent clustering. Hierarchical clustering is another common statistical method used to classify multiple samples into groups (clusters). The results are visualized as dendrograms. The length of an edge in a dendrogram between a cluster and its split is proportional to the dissimilarity (Euclidean distance) between the split clusters (Wittek, 2014). Dendrograms were used for two purposes in this study, to investigate the volatilomic similarities between the examined bacterial strains, and to determine what compounds were responsible for the discrimination of the bacterial volatilomes across the different media. Heatmaps coupled with dendrograms complement the PCA results (Figure 1) and demonstrate clear discrimination of the bacterial species in each of the examined media (Supplementary Figures 1–3). The heatmaps also offer minor insights at the strain-level differences that are present in the volatilomes as the samples from P. aeruginosa (Supplementary Figures 1, 2) are roughly clustered together. In each of the examined media, it appeared that the same compounds were responsible for the discrimination of the bacterial volatilomes (Supplementary Figures 1–3). Following the analyses, the chemical classes and compounds responsible for the observed media-, species- and strain-dependent differences were further investigated.

Although the chemical compositions of each bacterial species demonstrated a high degree of stability (Figure 2), the abundances of chemical classes varied across BHI, TSB and LB growth medium (Figure 2). The volatilomes of both E. coli strains were dominated primarily by the emission of indole and fatty alcohols (collectively ∼ 85%), with the remaining 15% of the volatilome being mostly acids and ketones (Figure 2). Indole is produced in a one-step reaction by the enzymatic catalysation of the amino acid tryptophan (Lemfack et al., 2020). Growth media free of glucose such as LB medium contain an abundance of amino acids (Sezonov et al., 2007) that can be alternatively metabolized to form indole which explains why indole abundances were relatively uniform across the media (Figure 2). E. coli primarily produces fatty alcohols via the fatty acid metabolic pathways (Liu et al., 2016), however, fatty alcohols can also be derived from glucose metabolism (Youngquist et al., 2013). A variety of fatty alcohols were identified at varying abundances across the examined media (Figure 3 and Supplementary Figure 5) between the two E. coli strains. Despite EC.B cells having a higher growth rate, EC.A emitted significantly higher abundances of 1-tetradecanol and 1-hexadecanol than EC.B across the media (Figure 3) and temporally (Figure 4). This supports the fact that the emission of these compounds is not just simply dependent on the progressive growth of cells, but rather the different metabolic pathways each strain utilizes as the cells proliferate. Fatty alcohols (C₃–C₁₆) have been frequently reported (van den Berg et al., 2006; Bos et al., 2013; Chen et al., 2016; Fitzgerald et al., 2020) to be emitted by E. coli with varying combinations being emitted strain-to-strain. Acetic acid abundances were high in both strains (Figure 3) and demonstrated a proportionate increase with cell proliferation in the first 5 h of incubation (Figure 4).

P. aeruginosa is a common gram-negative bacteria that has been associated with severe infections of burns (Lyczak et al., 2000) and diabetic foot ulcers (Abdulrazak et al., 2005), and it has also been previously labeled as the most common cause of persistent, fatal respiratory infections in cystic fibrosis patients (Bradbury, 2009; Davis et al., 2020). The chemical composition of the observed P. aeruginosa volatilomes was primarily made up of alcohols and hydrocarbons (∼75%) across the media (Figure 2). This appears to be in agreement with recent results published by Davis et al.(Scott-Thomas et al., 2010), although there are differences in the P. aeruginosa volatilomes at the compound-level. The remaining 25% of the volatilomes were composed of ketones, pyrroles and fatty acid esters (Figure 2). Characteristic compounds (Figure 3) emitted by both strains across the three media that have also been frequently detected in P. aeruginosa volatilomic analyses included 2-aminoacetophenone (Scott-Thomas et al., 2010; Shestivska et al., 2012; Tait et al., 2013), 2-undecanol (Filipiak et al., 2012b; Fitzgerald et al., 2020), 2-undecanone (Tait et al., 2013; Timm et al., 2018; Fitzgerald et al., 2020), 1-undecene (Filipiak et al., 2012b; Shestivska et al., 2012; Tait et al., 2013; Fitzgerald et al., 2020), 2-nonanone (Scott-Thomas et al., 2010; Filipiak et al., 2012b; Fitzgerald et al., 2020), and pyrrole (Filipiak et al., 2012b; Tait et al., 2013; Fitzgerald et al., 2020). Production of 1-undecene by both strains of P. aeruginosa was enhanced when the samples were cultured in LB medium most likely due to the absence of glucose (Figures 2, 3). Although significant differences in 1-undecene emission were observed between both strains across the media (Figure 3) and kinetically (Figure 4), 1-undecene was shown to be a temporal growth marker of both strains. This long chain alkene is derived from the metabolism of fatty acids and is a key component of the volatilomes of various Pseudomonas spp. (Dandurishvili et al., 2010; Kanchiswamy et al., 2015; Lo Cantore et al., 2015). 2-aminoacetophenone is produced in the amino acid degeneration (shikimate) pathway via the loss of a hydroxyl group on anthranilic acid (Weisskopf et al., 2021) [derived from chorismate via chorismate lyase (Parthasarathy et al., 2018)]. Despite being widely reported as a characteristic marker for P. aeruginosa (Shestivska et al., 2012), we observed a high degree of strain-dependent variation in both its emission across the different media (Figure 3) and kinetically (Figure 4). Our results suggest that the amino acid degradation pathway may potentially be more active in some P. aeruginosa strains than others. Differences in 2-aminoacetophenone production could also be due to strain-level differences in the regulation of quorum sensing (Kesarwani et al., 2011).

Since the emergence of antibiotic-resistant strains of S. aureus that caused epidemics in the 1950s and 1960s there have been large global efforts to develop early detection systems (Cookson, 2011). In the last 15 years, the understanding of the core and accessory components of the S. aureus volatilome has been steadily growing due to a growing number of in vitro volatilomic profiling studies (Bunge et al., 2008; Filipiak et al., 2012b; Baptista et al., 2019; Fitzgerald et al., 2020; Jenkins and Bean, 2020). The clinical value of in vitro volatilomic profiling has been demonstrated by Filipiak et al. (2015) who reported detection of VOCs, known to be emitted by S. aureus, in the breath of patients with S. aureus-positive respiratory infections. In this study here, the S. aureus volatilome demonstrated the highest degree of stability across the media out of the three examined species. This was illustrated by the tight clusters of “SA” samples in each score plot in Figure 1 and in the heatmaps (Supplementary Figures 1–3). Across TSB, BHI and LB media, the S. aureus volatilome was composed primarily of acids and alcohols (∼80% total volatilome), at different ratios in each. Other chemical classes recovered in all media included ketones and fatty acid esters. The characteristic compounds that were emitted by both S. aureus strains were not surprising and included 3-methylbutyric acid, acetic acid, ethyl 2-methylbutyrate and acetoin—all of which have been frequently reported in the literature (Filipiak et al., 2012b, 2015; Chen et al., 2016; Jenkins and Bean, 2020). These key compounds all originate from primary metabolic pathways (Weisskopf et al., 2021). 3-methylbutyric acid is derived from the metabolism of amino acids (leucine) while acetic acid, ethyl 2-methylbutyrate and acetoin arise from different stages of the fermentation process in glucose metabolism (Schulz et al., 2020; Weisskopf et al., 2021). The different primary metabolic pathways from which these compounds are derived are reflected in Figure 3, as in the glucose-free LB medium, acetic acid, ethyl 2-methylbutyrate and acetoin abundances are significantly lower than abundances observed in TSB and BHI while 3-methylbutyric acid emission was comparable across the media due to presence of amino acid substrates in all media. Similar reductive effects on the emission of acids and esters by S. aureus in glucose-free media was recently reported by Jenkins and Bean (2020).

Comprehensive bacterial volatilomic data will only be obtained through the analysis of a high number of strains. Strain-to-strain volatilomic variability has been relatively under studied, however, studies that have been carried out have highlighted potential strain-specific differences in bacterial volatilomes. Purcaro et al. (2019) reported discrimination of multiple strains of P. aeruginosa in the breath of infected mice. In our previous study (Fitzgerald et al., 2020), when analyzing triplicate samples of the same strains, we observed small differences between both P. aeruginosa and E. coli strains and concluded that further work is needed incorporating a higher number of samples to comprehensively resolve these differences. Analyzing different microbial strains at specific stages of cell growth and development has been recently proposed as an effective approach to comprehensively elucidate the variation in microbial volatilomes at the strain-level (Kai, 2020). In this study, we employed this approach in a more targeted manner to further investigate the strain-dependent emission variation of specific compounds—observed across multiple media—in P. aeruginosa and E. coli samples. The resulting kinetic plots shown in Figure 4 confirm that there was consistent strain-level specificity in the temporal emission of individual compounds. These results were in agreement with our earlier observed differences between the strains for the same compounds in Figure 3. The results obtained from this study also validate observed differences in VOC emission previously reported (Fitzgerald et al., 2020) between P. aeruginosa strains. Although the volatilomes obtained from all of the E. coli samples in this study are mostly in agreement with those previously reported (Fitzgerald et al., 2020), the same strain-level differences were not observed. It should also be noted, that when analyzing the full spectrum of VOCs recovered from these strains in an unsupervised manner via PCA (Figure 1), neither P. aeruginosa nor E. coli strains were discriminated from each other. Our results show that when investigating the full volatilomes of these strains, clear discrimination of the strains will not be achieved, it is only when the volatilomes are investigated at the compound level with corresponding emission kinetic data that these strain-level differences can be fully elucidated. While our data clearly shows strain-level volatilomic differences exist within specific species of bacteria, it also highlights the complexity of strain-level volatilomic discrimination. Strain-level differences were not observed between the examined S. aureus strains in this study, however, volatilomic discrimination of different S. aureus strains (MSSA and MRSA) has been previously reported (Boots et al., 2014) indicating that strain-dependent antibiotic sensitivities could also be potential factor in strain-level volatilomic diversity. Baptista et al. (2019) reported a high number of compounds and also demonstrated strain-level volatilomic differences across S. aureus samples based on enterotoxicity. Microbial volatilomic variability across different media has been frequently reported in the literature (Tait et al., 2013; Rees et al., 2017; Zareian et al., 2018; Jenkins and Bean, 2019). This was particularly highlighted by Rees et al. (2017) who detected a total of 365 compounds from 9 Klebsiella pneumoniae clinical isolates across TSB, LB, BHI and MHB media—only 10% of the compounds were common across all examined media. Many of the studies discussed in this text have highlighted the complexity and high degree of specificity of microbial volatilomes. Future work should focus on investigating the strain-level volatilomic diversity present within specific microbial species by analyzing a higher number of strains obtained from ecologically varying environments across different media. It is through, firstly, comprehensively investigating the factors that govern the emission of pathogen-specific metabolites that will allow volatilomics to be employed in clinics in the future.

Bacterial volatilomes are influenced by different nutritional environments and strain-level differences. By investigating this stability in this work a comprehensive understanding of the volatilomes of the examined bacterial species was achieved. Our objectives in this study were: (1) to obtain comprehensive volatilomic data from multiple strains of the wound-associated pathogens S. aureus, P. aeruginosa and E. coli in three different growth media; (2) to assess the stability and variation of the observed bacterial volatilomes; and (3) to assess strain-level specificity within the examined strains by comparing the emission of specific compounds at progressive stages of growth and development of the cells. Using HS-SPME GC-MS analysis we successfully analyzed the VOCs produced from each strain to obtain characteristic species-specific volatilomes. The observed volatilomes demonstrated a high degree of stability across the examined media, however, glucose-free media had a reductive effect on the emission of various primary metabolites. Strain-level variation was observed in P. aeruginosa and E. coli samples across the examined media in the emission of particular compounds. Comparative temporal volatilomic analysis of these strains confirmed that there were differences in the emission of individual compounds between the examined strains. Moving forward in microbial volatilomics, performing similar multi-strain kinetic experiments will provide a more comprehensive view of the capabilities of microbial volatilomes. Additionally, analyzing microbial volatilomes in different growth media allows specific metabolic pathways responsible for VOC production to be intimately investigated. Building a comprehensive understanding of the limits and possibilities of microbial volatilomics will elucidate what can be ultimately achieved in future applications.

On a clinical level, these pathogens pose a significant challenge as they cause severe wound infections. We are currently working on the study of the diabetic foot ulcer volatilome and have seen that some characteristic compounds (e.g., 3-methyl-1-butanol, 3-methylbutyric acid, and ethanol) also persist in wound samples. Our clinical work is still in the early stages, however, we hope to identify infection-specific volatilomic patterns that will potentially allow the detection of wound infections in the early stages.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://figshare.com/articles/dataset/Wound-associated_bacterial_pathogens_volatilomic_data/16692217.

SF performed the experimental work and drafted the manuscript. LH and AM supervised the work and reviewed the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.